Bob Young (@Bobby___Young) header
Bob Young (@Bobby___Young) profile picture

Builder from @Tsinghua_Uni/@Kimi_Moonshot Angel-backed by F&F Capital (pre-product) Former CEO(Chief Erhu Officer🎶) of a band

Joined January 2022
Photo by Bob Young (@Bobby___Young)Photo by Bob Young (@Bobby___Young)Photo by Bob Young (@Bobby___Young)Photo by Bob Young (@Bobby___Young)Photo by Bob Young (@Bobby___Young)Photo by Bob Young (@Bobby___Young)Photo by Bob Young (@Bobby___Young)Photo by Bob Young (@Bobby___Young)Photo by Bob Young (@Bobby___Young)Photo by Bob Young (@Bobby___Young)
再次感谢这篇文章的两份重要参考: 来自@haoailab的 vstack 以及@yenwod_ 关于HFT非常简洁、精要的文章,再次推荐: teapawt.org/yenwod_/status/2119896… 我也在迭代我的套利框架其他side-project,如果有新的进展再继续分享吧
Inspired by recent discussion about whether it's possible to learn trading outside an elite firm. If you're the kind of person that reads this project and thinks it sounds fun, imo you'll be able to learn employable trading skills just fine on your own.
3
1
7
1,141
Great work from @haoailab on vstack. Bridges and financial systems face the same problem: end-to-end testing is extremely expensive. With formal methods, we can specify the assumptions (A) and the requirements (P), then find a design (M) such that A ∧ M ⇒ P. Once this is proven, we only need to test the external interfaces and configurations. I’m applying this approach to a cross-chain bridge here: github.com/YoungY620/verifie… This opens the door to designs that go beyond human cognitive load.
Towards Verifiable LLM Infra 🚀 Suppose coding agents can build a fast inference engine from scratch, and one engineer orchestrates 10,000 agents in parallel: how do we know the engine they wrote is actually correct? Having agents review other agents mostly wastes tokens compared to empirical and formal verification. We introduce vstack, a 5-stage formal verification workflow for coding agents, and veriserve, a verifiable PD-disaggregated serving engine: uncovering 3 classes of real RDMA and cache races in production serving engines, and backing veriserve's KV-page lifecycle design model with a 226-module, 516,973-obligation TLAPS safety proof at +0.18% mean p50 TPOT overhead on GB200. Blog: haoailab.com/blogs/verifiabl… veriserve: github.com/hao-ai-lab/verise… vstack: github.com/hao-ai-lab/vstack
103
“要写自己想读的论文,就要做自己感兴趣的研究。”
Because I'm hiring a bunch of new PhD students, I took some time to think about what it means to start a research career in the age of competent AI agents. And then I wrote another long blog post. I'm not pompous enough to call it an essay, but you may want to read it anyway.
1
1
91
开会前先和kimi聊清楚是好文明
1
3
152
最近沉迷于制作表情包无法自拔
2
2
237
她不一样,她懂显卡。
1
144
Some footnote needed. If you can clearly state the question, you solve the question. The tricky thing is not the stochasticity of the LLM itself, but that of the world we expect it to formalize. Great thanks to the inspiring sharing from @halvarflake. Some humble additions here.🫡
88
最近的感觉就如图,感觉没有什么特别想输出和钻研的。 感受到身边的朋友同事也有类似的情绪。聊的都是游戏和八卦,和去年年底大不相同了。 消灭了软件的成本,也消灭了软件的需求。 当生产力拉到极限,继续拉动盈利的就是触达和获客了。 触达和获客?一瞬间大家都需要靠 争奇斗艳、扭捏作态 来博取关注? 工程师们十指不沾阳春水,最讨厌干这种活了🤣 北京终于天晴了,该出去走走了
3
202
感谢老师认可😇 给这段文字点赞也很符合年级第一的风范了hh😜
推上有个人形容 @ryolu_ 就像是一堆广告里播了一集电视剧。 你知道这种 “松弛感” 像什么吗?就像小时候年级第一上台表演才艺,让我想到了马太效应、优绩主义和自证预言。 几个月前 ryo 来 Kimi 被问:“对Kimi的印象快问快答” 答曰:“Distallation” 依我看,Distallation是这个公司最小的罪过了。
1
153
推上有个人形容 @ryolu_ 就像是一堆广告里播了一集电视剧。 你知道这种 “松弛感” 像什么吗?就像小时候年级第一上台表演才艺,让我想到了马太效应、优绩主义和自证预言。 几个月前 ryo 来 Kimi 被问:“对Kimi的印象快问快答” 答曰:“Distallation” 依我看,Distallation是这个公司最小的罪过了。
So here it is — @ryolu_ 's talk at @aiDotEngineer 2026 Shanghai. I shot it handheld on a Xiaomi 15 Ultra, so there's some shake; I fixed as much of it as I could in CapCut. Handheld footage just isn't going to be sharp or steady. Still, it's the best talk I saw on day 1 of AIE — which is why I'm not posting a text summary. Watch for yourself. I missed the first minute or so. Ryo opens with the excitement of what AI makes possible, and how seeing what others are building can quickly turn that excitement into anxiety about falling behind.
3
1
11
4,607
突然想到等到下一版模型编号为Kimi K49 那个重出江湖的男人 还是个和本体完全不同的克隆体 自嘲拉满了
1
137
我再说一遍,丢钱的是cex,据目前我所知可能的原因和密码学没有任何关系 看来 Web3 还是有希望的,这韭菜还没死绝啊
币安这次巨额失窃不是说明币安的安全建设不到位,而是 在AI 时代基于传统的安全策略都很脆弱。比特币算法在传统数学理论上的确无解,但是基于人的习惯的大逻辑下,问题会收敛很多很多。
1
2
159
每次和朋友出来吃饭,都会被拷打关于“蒸”的问题 这个我真不知道😅 也许Nathan的人脉比我知道的多。 youtu.be/BKuK0M6oDTI?si=FsPg…
161
诶趁粉丝少:你们推崇的deepseek super-agent“论文”在我看来就是一坨 “说点大家不知道的😅” 但我又意识到真的有人是蒙在鼓里的: 几年前我还关注了“小林”看各种面经和总结,学到了很多,帮助我找到了实习 claude code源码泄露那天又看到小林更新了,很可惜在我看来他就是在胡说八道🤔 当年来kimi确实是对的
一个关键的 insight:如今的LLM RL已经不是早年RL的算法驱动结果,而是数据驱动结果了。一个明显的提现: 左图:“随着时间的推移,多个研究项目展现了 REINFORCE 类算法的潜力,这些算法因其比 PPO 更简洁而备受赞誉” 而我当年的笔记上(图2),REINFORCE都是作为被优化的对象,用作课程的引子的 很难再看到从前寸纸寸金的好论文,以及读懂论文的那种兴奋感了(图3当年还写了博客)
2
201
推荐这个app,用我的邀请码可以薅羊毛,也可以给我加金币🥳 或者点这个链接 link.pokerqz.com/x3rX/invite…
德扑好有意思,如果是真钱就好了 我发现好多教程都不够简洁,实际上只需要不犯错误就能收割疯子了 我觉得应该趁着没有知识诅咒写个总结
79
德扑好有意思,如果是真钱就好了 我发现好多教程都不够简洁,实际上只需要不犯错误就能收割疯子了 我觉得应该趁着没有知识诅咒写个总结
1
127